Genetic engineering has made it possible to produce large amounts of polypeptides encoded by cloned DNA by means of recombinant expression systems, especially by expression in such prokaryotes as Escherichia coli (E. coli). The expressed heterologous polypeptide, which would otherwise either not be produced at all by the host cell or be produced only in limited amounts, may constitute a significant proportion of the total cellular polypeptide of the host cell.
Several problems are frequently encountered, however. Polypeptides over-expressed in the bacterial cytoplasm often accumulate as insoluble "inclusion bodies" (Williams et al., Science 215:687-688, 1982; Schoner et al., Biotechnology 3:151-154, 1985). Inclusion body formation is not limited to bacterial expression systems. For example, the Kruippel gene product of Drosophila can form inclusion bodies when produced in insect cells using a baculovirus expression system. Polypeptides accumulated in the form of inclusion bodies are relatively useless for screening purposes in biological or biochemical assays, or as pharmaceutical agents. Conversion of this insoluble material into active, soluble polypeptide requires slow and difficult solubilization and refolding protocols which often greatly reduce the net yield of biologically active polypeptide.
Even when heterologous polypeptides arc expressed in the cytoplasm of bacteria in soluble form, they often accumulate poorly as a result of degradation by host proteases. Further, the accumulated polypeptides often have a different amino terminus than that which is desired.
One approach to these problems is to fuse a polypeptide of interest to a polypeptide fusion partner such as the lacZ and trpE gene products (Goeddel et al., Proc. Natl. Acad. Sci. USA. 76:106-110, 1979; Furman et al. Biotechnology 5:1047-1051, 1987); maltose-binding polypeptide (Di Guan et al., Gene 67:21-30, 1988); glutathione-S-transferase (Johnson, Nature 338:585-587, 1989); ubiquitin (Miller et al., Biotechnology 7:698-704, 1989); or thioredoxin (LaVallie et al., Biotechnology 11:187-193, 1993). Often the fusion partner confers such desirable characteristics as greater solubility on the polypeptide of interest, especially when the recombinant host is cultured at temperatures below the optimum for growth (LaVallie et al., 1993, op. cit.). Low-temperature culture, however, introduces other practical problems which may make the process less suitable on a commercial scale.
The use of polypeptide fusions also allows the production of polypeptides which might otherwise be too small to accumulate efficiently in the recombinant host (Schultz et al., J. Bacteriol. 169:5385-5392, 1987). Further, appropriate fusion partners may act, e.g., as affinity peptides, facilitating recovery and purification of the fusion polypeptide from cell extracts containing hundreds of other polypeptides (see, e.g., WO 91/11454).
The use of fusion polypeptides has drawbacks, however. It is often necessary to cleave the desired polypeptide away from the fusion partner by enzymatic or chemical means. This can be accomplished by placing an appropriate target sequence for cleavage between that for the fusion partner and for the desired polypeptide. Unfortunately, the enzymes most widely used for polypeptide cleavage are expensive, inefficient, or imprecise in their cleavage, and cannot always be successfully applied to a majority of fusion constructs. For example, while enterokinase and Factor Xa are highly specific endoproteases, these mammalian enzymes are expensive to produce and require that a polypeptide of interest expressed in a prokaryotic host cell be isolated from the host cell before being treated with the mammalian enzyme, adding considerable expense to a large-scale process. In addition, the efficiency and specificity with which some enzymes cleave substrates is highly variable. While an enzyme like subtilisin, for example, may be relatively inexpensive to produce, the precision with which it cleaves substrates is less than acceptable for commercial-scale processes under current "Good Manufacturing Practices" (GMP).
Some yeast ubiquitin hydrolases efficiently cleave fusions in which ubiquitin is the fusion partner and the amino acid immediately downstream of the cleavage site is not proline (Miller et al., op. cit., 1989; Tobias and Varshavsky, J. Biol. Chem. 266:12021-12028, 1991; see also WO 88/02406 and WO 89/09829). One ubiquitin hydrolase gene cloned from the yeast Saccharomyces cerevisiae (S. cerevisiae), YUH-1 (Miller et al., op. cit. 1989), will not efficiently cleave fusions in which the downstream polypeptide is larger than about 25 kD. Another S. cerevisiae ubiquitin hydrolase gene (Tobias and Varshavsky, J. Biol. Chem. 266:12021-12028, 1991) is capable of cleaving ubiquitin fusions in which the polypeptide downstream of the cleavage site is as large as 130 kD. Both ubiquitin hydrolases are active when expressed intracellularly in E. Coli, allowing them to be used to cleave fusions in vivo. However, the use of ubiquitin as a fusion partner is hampered by the fact that multi-copy plasmids carrying ubiquitin fusion constructs may cause E. coli host cells, for example, to grow slowly and lose viability.
Cytoplasmic accumulation of fusion polypeptides suffers from the drawback that the heterologous polypeptide moiety may not be able to fold correctly in the strong reducing environment of the cytoplasm, leading to poor yields of biologically active polypeptide. To overcome this problem the polypeptide of interest may be fused to a "signal peptide," a short (15-30 amino acid) sequence present at the amino terminus of precursor polypeptides destined for secretion, i.e., export to non-cytoplasmic locations. In E. coli such locations would include the inner membrane, periplasmic space, cell wall and outer membrane. Typically, at some point just prior to or during transport of polypeptides out of the cytoplasm, the signal sequence is removed by host enzymes to produce the "mature" polypeptide. In these cases in which the signal sequence is removed by host enzymes to produce the "mature" polypeptide, the signal sequences are also known as "leader peptides" (For a recent review of the general secretory pathway in gram-negative bacteria and a discussion of leader peptides, see Pugsley, Microbiol. Rev. 57:50-108, 1993).
Localization of an expressed polypeptide to the periplasmic space is advantageous because simpler methods of polypeptide recovery can be used, including "osmotic shock" and other techniques. Although leader sequences may be used to deliver heterologous polypeptides into the periplasmic space of E. coli, few polypeptides are efficiently accumulated in soluble form by this method. Translocation of polypeptides across the lipid bilayer of the inner membrane appears to be inefficient, particularly in the case of fusions comprising leader sequences linked to heterologous polypeptides.
Only a few polypeptides that naturally lack a leader sequence are secreted to non-cytoplasmic (or periplasmic) locations, as demonstrated by their selective release from cells upon treatment with osmotic shock or freeze-thaw protocols. These include thioredoxin (Lunn and Pigiet, op. cit., 1982) and elongation factor-Tu (EF-Tu) (Jacobson et al., Biochemistry 15:2297-2302, 1976). IL-1-.beta. expressed in E. coli has been extracted by a modified osmotic shock procedure (Joseph-Liauzun et al., op. cit., 1990).
Extracellular localization may also be advantageous and may be accomplished by at least two different strategies: (1) Permeabilization of the outer membrane, allowing periplasmic polypeptides to "leak" out (U.S. Pat. No. 4,595,658; Kato el al., Gene 54:197-202, 1987); and (2) fusion to sequences which direct extracellular export (Nagahari et al., EMBO J. 4:3589-3592, 1985; U.S. Pat. No. 5,143,830). However, these methods do not work in many cases; and even if they do work, the methods generally are inefficient and often do not produce polypeptides with the desired amino terminus (Holland et al., Biochimie 72:131-141, 1990).
In the construction of a fusion polypeptide, the ideal fusion partner would be one which is useful for the production of a wide variety of heterologous polypeptides in a recombinant host cell, e.g, E. coli, at optimum growth temperatures. Preferably, such a fusion partner would improve the accumulation of the desired polypeptide in soluble, active form in a cellular location in which it is protected, e.g., from proteolysis, and where the fusion polypeptide may be recovered by simplified procedures. It would also be advantageous if such a fusion partner would allow the use of an efficient, inexpensive and precise cleavage system in vivo.